20.07.2013 Views

Notes on computational linguistics.pdf - UCLA Department of ...

Notes on computational linguistics.pdf - UCLA Department of ...

Notes on computational linguistics.pdf - UCLA Department of ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Stabler - Lx 185/209 2003<br />

but also things like<br />

human(Socrates ∧ Plato)<br />

(human ∧ Greek)(Socrates)<br />

((snub-nosed ∧ Greek)(human))(Socrates)<br />

(((to ∧ from)(Athens))(walked))(Socrates)<br />

• Standard logical inference is deep, uses few inference rules, and depends <strong>on</strong> few premises, while typical<br />

human reas<strong>on</strong>ing seems rather shallow, with possibly a large number <strong>of</strong> inference rules and multiple<br />

supports for each premise. – We discuss this in §16.5.1 below.<br />

• Standard logical inference seems well designed for m<strong>on</strong>ot<strong>on</strong>icity-based inferences, and negative-polarity<br />

items <strong>of</strong> various kinds (any, ever, yet, a red cent, give a damn, <strong>on</strong>e bit, budge an inch) provide a visible<br />

syntactic reflex <strong>of</strong> this. For example:<br />

i. every publisher <strong>of</strong> any book will get his m<strong>on</strong>ey<br />

ii. * every publisher <strong>of</strong> Plato will get any m<strong>on</strong>ey<br />

iii. no publisher <strong>of</strong> Plato will get any m<strong>on</strong>ey<br />

We see in these sentences that the c<strong>on</strong>texts in which any can appear with this meaning depend <strong>on</strong> the<br />

quantifier in some way. Roughly, any can appear <strong>on</strong>ly in m<strong>on</strong>ot<strong>on</strong>e decreasing c<strong>on</strong>texts – where this<br />

noti<strong>on</strong> is explained below, a noti<strong>on</strong> that is relevant for a very powerful inference step. We will see that<br />

“the sec<strong>on</strong>d argument <strong>of</strong> every” is increasing, but “the sec<strong>on</strong>d argument <strong>of</strong> no” isdecreasing.<br />

12 Review: first semantic categories<br />

12.1 Things<br />

Let’s assume that we are talking about a certain domain, a certain collecti<strong>on</strong> <strong>of</strong> things. In a trivial case, we<br />

might be discussing just John and Mary, and so our domain <strong>of</strong> things, or entities is:<br />

E ={j,m}.<br />

A simple idea is that names like John refer to elements <strong>of</strong> the universe, but M<strong>on</strong>tague and Keenan and many<br />

others have argued against this idea. So we will also reject that idea and assume that no linguistic expressi<strong>on</strong>s<br />

refer directly to elements <strong>of</strong> E.<br />

12.2 Properties <strong>of</strong> things<br />

The denotati<strong>on</strong>s <strong>of</strong> unary predicates will be properties, which we will identify “extensi<strong>on</strong>ally,” as the sets <strong>of</strong><br />

things that have the properties. When E is the set above, there are <strong>on</strong>ly 4 different properties <strong>of</strong> things,<br />

℘(E) ={∅, {j}, {m}, {j,m}}.<br />

We can reveal some important relati<strong>on</strong>s am<strong>on</strong>g these by displaying them with with arcs indicating subset<br />

relati<strong>on</strong>s am<strong>on</strong>g them as follows:<br />

234

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!